JOURNAL
OF EXPERIMENTAL
SOCIAL
13,315332
PSYCHOLOGY
(1977)
Persuasive Argumentation
and Social Comparison
as Determinants of Attitude Polarization
EUGENE BURNSTEIN
The University
of Michigan
AND
AMIRAM
Tel-Aviv
VINOKUR
University
Several of our studies indicate that persuasive-arguments theory by itself is an
adequate explanation of polarization. Sanders and Baron (Journal of Experimental
Social Psychology,
1977,13,303-314) criticize this research. More generally, they
contend that both argumentation and comparison are involved, “with persuasive
arguments facilitating the shifts motivated by social comparison.” We feel that their
critique is unconvincing. Relevant portions of the standard literature are reviewed
to demonstrate that social comparison is neither a necessary nor sufficient
condition for polarization. Finally, we speculate about how persuasive-arguments
theory could be extended to argument-poor settings (e.g., Asch’s line comparison
situation).
Experience teaches us that the critical finding is a will-o’-the-wisp.
Theoretical disputes in social psychology are settled, if ever, by the overall
weight of evidence. So in the last analysis, whether persuasive
argumentation
alone or both persuasive argumentation
and social
comparison are needed to explain polarization will be decided on the basis
of all the relevant data, which are massive, and not just on the few studies
discussed by Sanders and Baron (1977). Our reading of the literature
suggests that informational
theories, such as persuasive-arguments,
describe the immediate causes of polarization and by themselves offer a
rough but reasonably complete account of the phenomenon.
Theories
based on normative influence, such as social comparison,
deal with
relatively remote processes which may lead to polarization on occasion,
when they are mediated by persuasive argumentation.
Sanders and Baron disagree with us, and they are not alone. In a recent
analysis of polarization research, Myers and Lamm (1976) conclude that
This paper was facilitated by Grant No. MH 29140-01 from the National Institute of Mental
Health. Requests for reprints should be sent to Eugene Bumstein, Research Center for Group
Dynamics, Institute for Social Research, University of Michigan, Ann Arbor Michigan 48106.
315
CopyrIght
All rlghfs
D 1977 by Academic Press. Inc.
of reproduction
in any form reserved.
ISSN 0022-1031
316
BURNSTEIN
AND
VINOKUR
“[although]
the evidence for [informational
theories such as persuasivearguments] is compelling . . . it also appears that group polarization is
not fully explained by a passive process of cognitive learning. More
dynamic processes of cognitive rehearsal and verbal commitment are also
likely contributors.”
Their basis for disagreement is different from that
stated in the Sanders and Baron paper. While we are concerned here mainly
with the latter, it should become clear that the processes described in
persuasive-arguments
theory are quite “dynamic,”
if not mercurial.
Indeed, it would be extremely difficult to analyze the construction and
comprehension of arguments without assuming that the person acts on and
transforms information (see also Anderson & Graesser, 1976).
THE PROBLEM
The basic idea of persuasive-arguments
theory is that when the person
evaluates (or re-evaluates) alternative X relative to alternative Y, he
generates arguments, namely, ideas, images, or thoughts describing the
attributes of X and Y. This process assumes that there exists a culturally
given pool of arguments speaking to each alternative. To judge the relative
merits of these alternatives, the person samples (retrieves arguments) from
this pool. Arguments may vary in availability
(the probability
of their
coming to mind), direction (pro-X and pro-Y), and persuasiveness. When
the preponderance of arguments in the pool favors a particular alternative,
the average prior attitude reflects the direction and magnitude of this
preponderance. Further thought or discussion leads to polarization toward
the alternative that initially elicits more and/or better arguments. Its
magnitude will depend on whether the initial argument samples (a) overlap
or (b) exhaust the larger pool. That is to say, polarization will be maximal
when a person begins to rethink the issue and many arguments remain that
have not yet come to mind, or when several individuals discuss the issue
with each other and not all of them have thought of the same arguments.
Let us illustrate persuasive-arguments
theory by a simple example.
Consider a choice in which the culturally given pool contains six pro-X
arguments, a, 6, c, d, e, andf, and three pro-Y arguments, 1, m, and n. If
three equally pro-X individuals discuss the issue, one of several distinct
outcomes would be predicted, depending on the distribution of arguments
among members. Say all three of our discussants had thought of the same
arguments. In this case, discussion would produce no change in their
attitudes toward X. On the other hand, if a, 6, and m had come to mind in
one; c, d, and m in the second; and e, A and m in the third (i.e., if each has
different pro-X arguments, but the same pro-Y arguments), then the
discussion would produce marked polarization
toward X. Finally,
polarization toward Y would be predicted if one member had generated a,
b, and 1, another a, 6, and m, and the third a, b, and n (i.e., if each had
ATTITUDE
POLARIZATION
317
initially
thought of the same pro-X but different pro-Y arguments).
Normally, individual argument samples are representative of the larger
pool. Therefore, average prediscussion preferences can be estimated from
the balance of pro-X and pro-Y arguments in the pool. Postdiscussion
preferences can also be predicted if, in addition, we know the degree of
overlap among individual argument samples (Vinokur & Burnstein, 1974).
An ingenious demonstration
of the latter point was made recently by
Kaplan and Miller (Note 2). They observed that when a series of arguments
is learned prior to discussion, recall of a particular argument will depend on
its position in the series (e.g., later arguments are remembered better than
earlier ones). Thus, there should be less overlap in argument samples (i.e.,
a greater variety of arguments will be recalled during discussion) and
consequently greater polarization when each member had learned the
arguments in a different order rather than in the same order. This was in fact
the case (also see Kaplan, Note 1, 1977).
Sanders and Baron propose that the information processes described by
persuasive-arguments
theory and the normative processes described by
social comparison theory have complementary
effects on polarization,
“with persuasive-arguments
facilitating
the shifts motivated by social
comparison.”
On balance, it seems to us that this proposal should be
rejected on two grounds. First, it offers a needlessly complicated
explanation of polarization. To illustrate, it is well known that polarization
and convergence occur simultaneously. (In fact there are several “neutral”
Choice-Dilemmas
that produce convergence without polarization.)
Specifically, the typical finding is that during discussion the most extreme
member moves to a relatively moderate position while the next most
extreme member hardly changes at all (e.g., Ferguson & Vidmar, 1971).
Although more than one version of social comparison theory is used to
explain polarization
(see below), all have difficulty in dealing with
convergence. Usually a second process is posited. This seems unnecessary
since persuasive-arguments
by itself accounts for both effects on the basis
of informational
influence. According to persuasive-arguments
theory,
attitude polarization
is fundamentally
an informational
phenomenon;
normative influences are relatively remote and operate on polarization, if at
all, through cognition. If a person is distressed or elated because others
hold attitudes more extreme than his own and he is about to shift toward
their position, this state of affairs will be reflected in ideation, that is, in the
content of the argument sample, and can be taken into account at that point.
In fact, we have observed somewhat to our surprise that when the attitudes
of others are sharply opposed to the ideas they express, the latter
determines the direction of polarization
(Burnstein & Vinokur,
1973,
Experiment II). Hence, for the purpose of explaining polarization,
social
comparison theory may be excess baggage. The Sanders and Baron
proposal is also difficult to justify in the light of past research. There is just
318
BURNSTEIN
AND
VINOKUR
too little evidence that social comparison processes have a direct affect on
polarization. This large literature has been diligently reviewed by Myers
and Lamm (1976). Thus, our discussion will focus on evidence that they
touched on either lightly or not at all.
PAST RESEARCH
Social comparison theory, in one version or another (see the review by
Pruitt, 1971), assumes: (a) a preference for alternative X is more socially
desirable than a preference for alternative Y; (b) the person believes his
own preference for X is at least as extreme as those of his peers (in Brown,
1965; Festinger, 1954; Jellison & Riskind, 1970) or is too extreme to be
acceptable (in Levinger & Schneider, 1969; Pruitt, 1971); (c) upon learning
this is untrue, he experiences distress (in the Brown, Festinger, and Jellison
& Riskind version) or relief (in the Levinger & Schneider and Pruitt
version); (d) either affective state causes the person to take a more extreme
position which results in a decrease in distress (e.g., because according to
Jellison & Riskind he no longer appears less able than others) or an increase
in satisfaction (e.g., because according to Pruitt he freely vents what was
formally suppressed). These assumptions as well as those of persuasivearguments theory predict the conditions under which polarization will
occur. Since these predictions often disagree, they can be pitted against
each other in experiments. When this was done, we found that (i) if an
individual
could argue but not compare, polarization
still occurred
(Burnstein & Vinokur, 1973); (ii) if he could compare but not openly argue,
polarization vanished (Burnstein & Vinokur, 1973, 1975) or was greatly
attentuated (Burnstein,
Vinokur,
& Trope, 1973); but (iii) even in
attentuated
form, polarization
seemed to depend directly on tacit
argumentation
(Burnstein & Vinokur,
1975). The latter experiment,
however, did suggest that comparison processes can indirectly influence
polarization;
that is, information about others may guide the person in
generating arguments. Sanders and Baron criticize each of our studies in
some detail. Therefore, it will be convenient to discuss these results later
when we reply to specifics of this critique. However, other findings also
warrant attention.
In an unpublished study, we once instructed each subject to predict the
distribution of choices of “100 people like you” over the response scale.
On the average subjects thought that about 30% of these persons would be
more extreme than themselves. The assumption that an individual will be
surprised, piqued, or elated to discover that one or two members of a
five-person group prefer a more drastic course than he therefore seems
unconvincing. Indeed, there is absolutely no evidence of a change in affect
either when such a discovery is made or when the individual shifts toward a
more extreme position. Yet these propositions about affective changes are
ATTITUDE
POLARIZATION
319
easily tested. Hence, the absence of published research on this point is
worrisome, especially since we do know that when the individual shifts
toward a more extreme position he does not think that his posterior
preference is any more socially desirable or adheres any more closely to
widely cherished standards of conduct than his relatively moderate prior
preference (Kahan, 1975). Moreover, diligent attempts to relate polarization to the belief that one is more extreme than his peers have been
singularly unsuccessful (e.g., Lamm, Trommsdorff,
& Rost-Schaude,
1972). Our own concern about the issue led us to ask subjects on a
postexperimental
questionnaire
to describe their state of mind upon
learning that their preferences differed from those of other members
(Burnstein & Vinokur, 1975). Reports of distress or relief were rare and
unrelated to polarization. However, polarization was correlated with the
extent to which the person tried to explain the difference and with the
amount of thought given to reasons why others made the choice they did (as
well as with the number of such reasons he actually generated). Although
the mental processes these people described are not those one would
anticipate on the basis of a social comparison analysis, they are just what
would be expected on the basis of persuasive-arguments
theory.
A second line of evidence comes from several experiments in which
members make their initial choices in terms of one feature of the decision
task and then discuss either the same or another feature. For example, in
testing a subjective expected utility (SEU) model with Choice-Dilemma
items (the most widely used decision-task
in polarization
research),
individuals indicated either (a) the minimal acceptable probability
of
success on the uncertain alternative or (b) the utility of the outcome
associated with the certain alternative. In some conditions, both initial
preferences and discussion had to do with the same feature (probability or
utility); in other conditions, they involved different features. An individual
ought to have a more clear-cut basis for comparing himself with others
when preferences and discussion involve similar rather than different
features. Thus, if social comparison processes are important, polarization
should be attenuated under the latter conditions. In fact, no attenuation
was observed (Burnstein,
Miller, Vinokur, Katz, & Crowley, 1971;
DesJarlais, 1971; Vinokur, 1971). Furthermore, the SEU model assumes
that extreme preferences are no more attractive per se than moderate ones
(e.g., there is no utility for risk). Social comparison theory assumes that
extreme preferences are more attractive. If the latter assumption were
correct, SEU maximization
would have been unable to account for the
data. The fit between predicted and obtained preference, however, was
extremely close, indicating that the social comparison assumption is
unnecessary.
Finally, two sets of findings suggest most cogently that social
comparison processes have little direct effect on polarization.
The first
320
BURNSTEIN
AND
VINOKUR
concerns what the person thinks about and what he says in the polarization
situation. Using the Choice-Dilemmas,
Vinokur, Trope, and Burnstein
(1975) performed a content analysis of the ideas that come to mind while a
person formulates his attitude as well as those expressed during discussion.
It was possible to identify several classes of arguments having to do with (a)
the perceived utility of success and failure associated with the uncertain
alternative and (b) the perceived utility of the certain alternative, both of
which were called oufcome utilities. In addition, since social comparison
theory assumes that an extreme choice is in itself attractive, utilities
associated with taking the “risky”
or “cautious”
action should exist
independent of the utilities associated with the outcomes actually specified
in Choice-Dilemma
items. Arguments pertaining to these action utilities
were also identified. (To illustrate, consider the choice between going to a
highly prestigious university where a good proportion flunk out and a
mediocre one where all receive their degree. In the face of this dilemma,
two common arguments regarding outcome utilities are “he can always
transfer to the easy school” and “he ought to be sure of getting his degree,”
the former, of course, supporting the choice of the uncertain alternative,
the latter supporting the certain alternative.
Examples of arguments
concerned with action utilities are “life is no fun unless you take a chance”
or “being cautious is a sign of wisdom.“)
In fact, over 70% of the arguments generated by subjects in the above
study (either privately or during discussion) were concerned with outcome
utilities, whereas less than 10% were concerned with action utilities.
Moreover, whereas outcome utilities were significantly correlated with
polarization (r = .63), action utilities were not (r = .16). Finally, when
outcome utilities (the persuasive arguments variable) and action utilities
(the social comparison variable) were combined to predict polarization, the
correlation was no higher (r = .62) than when only outcome utilities were
considered. If what a person thinks about, as well as what he says, reflects
the relative significance of argumentation
and social comparison, then
these results suggest that the former process by itself has considerable
impact on polarization, whereas the latter has very little.
While Sanders and Baron are less critical of the findings by Tesser and his
co-workers than of ours (specifically, Burnstein & Vinokur, 1975), both
sets of results raise the same vexing problems (Sadler & Tesser, 1973;
Tesser & Conlee, 1975; Tesser & Cowan, in press; Tesser and Leone,
1976). Using a variety of attitude objects, the Tesser group discovered that
if a moderately pro-X individual simply thinks about the issue, this is
sufficient to polarize his attitude toward X. In no instance did these effects
depend on knowledge about the attitude of others; they occur without being
“motivated
by social comparison.”
In addition,
the magnitude of
polarization was shown to vary positively with (i) the number of arguments
available to the person or his knowledge about the issue, (ii) the length of
ATJXTUDE
POLARIZATION
321
time he spends thinking about it, and negatively with (iii) the extent to
which he is distracted from thinking. These three variables have special
significance: In terms of social comparison theory, their relationships to
polarization makes no sense at all. It is self-evident, however, that each in a
different fashion govern the number of arguments likely to come to mind
and that each operates just as persuasive-arguments
theory would predict.
Our distraction experiment (Bumstein & Vinokur, 1975) is similar to
Tesser and Leone’s (1976) except that in the former the person knew the
preferences of others and in the latter he did not. Tesser used the
probability of polarization rather than the amount of polarization as his
dependent measure. When the comparable values are computed in our
study, it appears that if the person was not distracted from generating
arguments, the probability of polarization we observed (.48) was no greater
than that observed by Tesser and Leone (.53), despite the fact that social
comparison was possible in the former study but not in the latter.
Furthermore, both experiments demonstrated that polarization could be
made to decrease. Tesser and Leone found that this occurred to a similar
degree when persuasive arguments were unavailable to the person
(probability
of polarization = .33) and when he was prevented from
generating them (probability of polarization = .25). Moreover, the latter
value is comparable to that obtained with the Bumstein and Vinokur
distraction procedure (. 19) where the individual was perfectly aware that
others were more extreme than himself. These findings suggest (1) that
social comparison does little to enhance or sustain polarization, (2) that
polarization depends on a capacity to generate persuasive arguments, and
(3) that this capacity is diminished when the person either is presented with
an issue about which he has little knowledge or is prevented from thinking
about the issue.
The polarizing effect of attention and thought on individual judgments
has been observed in several different contexts. For instance, a large
literature indicates that evaluations of visual or linguistic stimuli polarize as
the frequency of exposure (and, thus, attention) to this material increases
(e.g., Brickman,
Redfield, Harrison, & Crandall, 1977; Gush, 1976;
Perlman & Oscamp, 1971; Zajonc, 1968). Especially relevant are the
studies comparing discussions and familiarization.
Some years ago
Bateson (1966) and Flanders and Thistlethwaite (1967) found that attitudes
also polarized when individuals become familiar with the issue. Soon after,
however, other studies appeared that were unable to obtain this effect (Bell
& Jamieson, 1970; Stokes, 1971; Teger, Pruitt, St. Jean, & Haaland, 1970).
The familiarization
procedure essentially consists of having the subject
think about and list arguments. On the face of it, this method seems quite
similar to that of Tesser. The fact that Tesser has repeatedly found
polarization to increase with thought, however, implies there must be some
divergence in procedure heretofore overlooked. Not all of these studies
322
BURNSTEIN
AND
VINOKUR
provide the methodological
information required for such an analysis.
Nevertheless, an examination of those that do suggests two consistent and
theoretically
significant differences; namely, the ins?ructional
set for
generating arguments and the time allowed for doing so. In Stokes (1971) as
well as in Bell and Jamieson (1970) the instructions stress objectivity,
thinking of arguments on “both sides” of the issue. By contrast, Bateson
(1966), Flanders and Thistlethwaite
(1967), and Tesser (1976) asked
subjects to “use (their) imagination,”
to “make sure (they have) not
overlooked some relevant information,”
to “read between the lines,” or
simply to “think about the issue until told to stop.” Our conjecture is that
the former instructions more so than the latter set the person to generate
arguments which arrest polarization (e.g., an equal number of pro-X and
pro-Y arguments).
The second difference in procedure has to do with the time made
available to the person for thinking about the issue before actually
expressing a preference. If the person has a long time to generate
arguments before indicating his initial position, he will exhaust the pool of
available arguments. When he is again given the opportunity to think about
the issue, no new arguments come to mind. As a result, his posterior
attitude remains unchanged. More generally, polarization should decrease
as the amount of thought given to a prior preference increases relative to
the amount given to a posterior preference. Both in Bateson and in Tesser,
the initial interval seems comparatively short. Bateson indicates that his
subjects had about five times longer to think about their posterior position
than their prior position. Tesser systematically
varied the posterior
interval. His initial interval, however, was always quite short; the subject
had to give his “immediate
reaction to the statement.”
Flanders and
Thistlethwaite do not describe the relative length of these intervals in their
experiment, nor do the studies which fail to obtain polarization, except for
Bell and Jamieson (1970) who do say no time limit was imposed on the
initial response. Presumably, this means the initial interval was determined
by the slowest subject in the groups and, thus, on the average, was
relatively long.
A REPLY TO CRITICISM
OF SPECIFIC
EXPERIMENTS
We feel that Sanders and Baron’s criticisms of several of our studies are
not well-founded.
To begin with, their objections to many of our
conclusions are based mainly on comparison of selected calls despite the
absence of reliable main effects. For example, in our original study
(Experiment I in Burnstein & Vinokur, 1973), four of the five discussion
items polarized (the fifth item approached but did not quite reach traditional
levels of statistical significance). In the unpublished
replication
that
Sanders and Baron cite, only one of the five polarized. Hence, a failure to
ATTITUDE
POLARIZATION
323
replicate would be the most appropriate claim, and criticism of our study
based upon the results pertaining to the single deviant item seems
unjustified.
Furthermore,
as Sanders and Baron note, the procedure used in this
replication
differed from the original (Experiment
I in Burnstein &
Vinokur,
1973) in at least one significant respect. The earlier study
attempted to demonstrate that persuasive argumentation was sufficient to
produce a shift and that knowledge of others’ actual attitudes was
unnecessary. It found that when a member does not know whether others
are arguing for their true position or its opposite, and the former in fact was
the case, typical polarization effects are obtained. Subjects were not asked
to infer the true position of others until the experiment was completed.
Sanders and Baron, however, repeatedly asked subjects to draw such
inference during the experiment. This procedure makes the inference task
more salient than in the standard polarization study. Thus, it may distract
subjects from thinking about the issue being discussed. In any case,
persistent prodding does probably provide a better estimate of what the
person is capable of inferring than our postexperimental
questionnaire.
Whether or not it is a more valid estimate of what he actually infers in the
typical polarization situation is problematic.
In our study (Burnstein &
Vinokur, 1973) we tried to create a situation in which inferences about
another’s attitude were not possible and the impossibility would be clear to
subjects. Whatever may have been the case, when subjects are not
repeatedly instructed to report the positions of others, we were able to
determine that such inferences had no effect on posterior choices:
Internal analysis of the shifts produced by the subjects who correctly inferred the
relationship between initial and advocated choice (a situation comparable to that of
subjects in the typical polarization study) showed that they were not different from
the shifts produced by the other subjects. In all cases the shifts of the correct
subjects were in the same direction as the shifts of the other subjects, and in no case
was the difference statistically significant. Indeed,
. contrary to what would be
predicted by comparison theories, the shifts of the correct subjects were smaller in
magnitude than the shifts of the other subjects (Burnstein & Vinokur, 1973; p. 129).
Internal analysis based on post facto questionnaire
responses are
inevitably ambiguous. This stricture applies to us as well as to Sanders and
Baron in their replication. There is a difference, however. The internal
analysis in Bumstein
and Vinokur
serves to emphasize that our
conclusions do not depend on the person being able to guess correctly the
position of others despite the experimental
manipulation.
But the
conclusions themselves are based on main effects. Nevertheless, in order
to be more certain about the consequence of guessing the position of
others, we must await an experiment in which this is systematically varied.
Still, it should be noted that several years ago a flurry of studies were
reported in which group members learned about each other’s position and
324
BURNSTEIN
AND VINOKUR
then indicated their posterior preference without discussion. There was no
need to guess or infer. If polarization occurred under these conditions,
social comparison must be a sufficient cause. In many cases (Bell &
Jamieson, 1970; Clark, Crockett, &Archer, 1971; St. Jean, 1970; St. Jean &
Percival, Note 3; Wallach & Kogan, 1965) no polarization was observed.
The remainder (Clark & Willems, 1969; Teger & Pruitt, 1967; Stokes, 1971)
found an attenuated effect. These results were among the first to suggest
that social comparison was not a sufficient cause and that an important
mediating process was being overlooked.
Experiment II ofthis same article (Burnstein & Vinokur, 1973) suggested
that if a member had to argue for the “mirror-image”
of his initial
position-thus
it should be difficult to muster persuasive argumentspolarization
will not occur, even though the “mirror-image”
rule was
known and another’s real attitude could be inferred. In fact, no risky item
polarized; only when combined did they evidence a marginally significant
shift toward risk. Similarly, no cautious item polarized; when combined,
however, a reliable but atypical shift (toward risk) appeared. It was
concluded that knowledge of others’ attitudes was insufficient to motivate
the typical polarization effect. Sanders and Baron say these results are
“open to the alternative hypothesis (that) the weak and atypical shifts (the
former referring to the marginal effect with risky items, the latter to the
more robust effect with cautious items) . . . were due to pitting two
active processes (argumentation
and comparison) against one another.”
Let us follow their “pitting against*’ hypothesis to its logical conclusion.
Our subjects were initially somewhat risky on the risky items and, given the
“mirror-image”
rule, tended to have to argue for caution. Because there
was a whiff of polarization toward risk, Sanders and Baron must assume
that comparison processes were slightly stronger. On cautious items these
same subjects had to argue for risk. Here, an atypical shift toward the
uncertain (risky) alternative was observed. If the “pitting
against”
hypothesis is to account for this effect, another assumption must now be
made which is exactly the opposite of the one needed to explain marginal
polarization on the risky item; namely, on the cautious items persuasive
argumentation was stronger than comparison processes. Such reasoning is
unconvincing.’
In Burnstein, Vinokur, and Trope (1973) individuals learned about
choices of others and arguments for these choices before indicating their
own preference. Both the number of choices and the number of arguments
were varied orthogonally
so that individuals in one condition received
many choices and few arguments: in another condition, few choices and
1 In criticizing this study, Sanders and Baron point out that self-persuasion through
role-playing could have contributed to the results obtained. However, to the extent that such
self-persuasion involves the self-generation of persuasive arguments, this possibility is not
inconsistent with our position.
ATTITUDE
POLARIZATION
325
many arguments, and so forth. The number of arguments had a reliable
effect, but the number of choices did not. That polarization varied only with
the number of arguments would seem to pose a real problem for social
comparison theory. The pattern of shifts over conditions, however,
suggested a cautious interpretation and led us to muse about how another’s
choice could, by stimulating thought, produce additional argumentation
without group discussion: “Knowledge that other’s choices are discrepant
from his own may induce the person to reconstruct a line of reasoning
(generate persuasive arguments) which he thinks could have produced (and
thus would support) such choices” (Bumstein, Vinokur, & Trope, 1973; p.
244). This conjecture was then tested in a later study (Burnstein & Vinokur,
1975) which Sanders and Baron also find wanting.2
Their criticism, however, brings us back to the heart of the matter. Is the
knowledge that one’s own preference is relatively moderate sufficient to
produce polarization? Or is there a more immediate cause, a process which
intervenes between this knowledge and attitude change? Bumstein and
Vinokur compared polarization across three conditions. In one, the person
learned exactly how his position differed from those of others and was then
encouraged to think further about the issue in question as well as to list all
arguments which came to mind. In the second, the person learned about the
difference between himself and others but was then distracted from
thinking further about the issue. The third condition was identical to the
first except the person received no information about another’s position.
Polarization occurred only in the first condition, its magnitude depending
on the balance of arguments actually generated. We concluded that
learning one was different is nor sufficient to produce a shift in attitude; this
knowledge was effective only to the extent that it moved the person to
construct arguments explaining the difference. These findings also partly
convinced Sanders and Baron. In their discussion of our study they agree
that recognizing one’s relative moderation is not enough, that there is a
more immediate cause involved. Polarization they say, will not occur when
the “cognitive work” which mediates the effect is blocked, as was the case
* In this context Sanders and Baron cite two interesting gambling experiments by
Blascovich and colleagues (Blascovich & Ginsburg, 1973; Blascovich, Ginsburg, & Veach,
1975) in which no differences in bets are obtained between interacting (argumentation plus
comparison) and coacting (comparison only) groups. In mentioning these findings they
correctly note that persuasive-arguments predicts no polarization when there is nothing to
discuss, namely, when it is impossible for the person to generate ideas. Sanders and Baron go
on to assume that in the Blascovich situation arguments must be “sparse.” According to
persuasive-arguments, polarization should not occur. But polarization did occur. They
conclude, therefore, it must be due to social comparison. To evaluate this line of reasoning&
of Blascovich’s major findings should be considered. The most important one is that
appreciable polarization, the extraordinary increase in the size of bets, occurred under
“alone” conditions, before the subjects had any opportunity to compare or to discuss. Ante
hoc ergo propter hoc?
326
BURNSTEINANDVINOKUR
in our distraction condition. The mediating processes that Sanders and
Baron suggest, however, are on the spur of the moment (e.g., “a desire not
to look wishy-washy, which requires them to engage in rationalization
before changing their position”)
and without theoretical or empirical
support in the polarization
literature.
The processes described by
persuasive-arguments
theory, on the other hand, are reasonably well
confirmed.
EXTENDING
PERSUASIVE-ARGUMENTS
THEORY
Two problems engage us at present. The first, which will be mentioned
only briefly, has to do with persuasion. Up to now a persuasive argument
has been defined simply as a statement judged to have a certain cogency.
Recently we began to consider the theoretical basis for persuasiveness and
ventured that this property depends on an argument’s validity and
originality.
According to this formulation,
an argument is an assertion or
inference which follows from a line of reasoning. Validity would then refer
to the soundness of the assertion-is
it justified? The line of reasoning,
which itself may involve a string of inferences, has as its premise the
information available to the person when he begins to think about the issue.
Ordinarily this preliminary information is contained in the way the issue is
stated to the person, what he knows about the positions of others, and so
forth. An argument’s originality is defined by the number of steps in aline of
reasoning. This is a rather important property since we assume that at each
step, from premise to assertion, additional ideas are recruited and made
available to the person. Therefore, when one either generates
or
comprehends
a valid argument, the greater its originality,
the more
additional arguments are brought to mind. So far it has been possible to
demonstrate crudely that if an argument is valid, both perceived and actual
persuasiveness increase with originality (Vinokur & Burnstein, in press).
We do not mean to imply that a valid line of reasoning must follow rules
of formal logic. Sometimes, however, either little information is available
with which to think about or discuss the choice (e.g., in Sheriffs
autokinetic studies) or the information available appears to argue for a
choice different from the one actually observed (e.g., in Asch’s line
judgment studies). Common sense might then suggest that informational
processes are irrelevant. Nevertheless, there may be a mental representation of a line of reasoning even though one has difficulty describing the
information it contains in words or images (see Natsoulas, 1970; Pylyshyn,
1973). Certainly it can be assumed that tacit argumentation occurs but is
difficult to tap. One must then make derivations on the basis of this
assumption that do have observable consequences. Which brings us
somewhat awkardly to our second research problem: Can persuasivearguments theory be applied in situations where, on the face of it,
Al-l-ITUDE
POLARIZATION
327
argumentation seems gravely enfeebled? We will be especially concerned
with cases in which normative influence is commonly thought to dominate
information influence.
Sanders and Baron believe that the autokinetic judgment study in Baron
and Roper (1976) is such a case. They reason that in this situation
informational processes are virtually nonexistent because pithy assertions
justifying a particular amount of movement seem unimaginable.
Shifts in
judgment, therefore, cannot be explained by argumentation and must be
due to normative (comparison) processes. On the other hand, it is at least as
plausible that by informing subjects that intelligent people give larger
estimates of movement (their procedure for making the latter social
desirable), Baron and Roper also provide them with a good argument for
making such estimates: “I am intelligent, intelligent people make large
estimates, therefore . . .” or “Intelligent
people tend to be correct, I want
to be correct, therefore. . . .” Moreover,
there are few persuasive
arguments to the contrary since the actual amount of movement is totally
ambiguous.
For our purposes the Asch line comparison
procedure is more
appropriate (Asch, 1956). The issue involved is completely unambiguous.
In this case, one might expect arguments supporting an inaccurate
judgment to be totally unconvincing and rare and those supporting the
accurate judgment to be only slightly more numerous but immediately
persuasive. We do not believe that undergraduates are so uncreative. In
fact, the postexperimental
interviews (Asch, 1956; Tuddenham
&
McBride, 1959) reveal a surprisingly large number of not unpersuasive and
even novel arguments which subjects attest to having affected their
decision either to conform or to dissent (e.g., “I figured the lines contained
some sort of illusion that I was not subject to and they were” or “I wanted
to make . . . (the experimenter’s)
results better . . . (to avoid) the
graph of results with a big dip in it.“). Whether or not the rich ideation of
Asch’s subjects belies the common sense view that the line judgment task is
“unthinkable”
(and surely “undiscussable”),
assume for the moment
argumentation occurs. Let us now show that this assumption has testable
implications
that are, on occasion, counterintuitive.
In this context a person’s prior and posterior choices of a comparison line
refer to the judgments he makes before and after he is exposed to the
preferences of others. The likelihood
of a shift toward inaccuracy (or
accuracy) depends on new ideas coming to mind during this interval.
Persuasive-arguments
theory assumes that whether a person’s prior
preference is strengthened, weakened, or remains unchanged, hinges on
the degree to which (a) the argument he generated when making his original
choice and (b) those he generates when induced to reconsider this choice
are representative
of the initial argument pool. If both are equally
representative, then reconsideration will have no effect. If, however, a
328
BURNSTEIN
AND
VINOKUR
posterior sample is less representative than the prior one, preferences are
likely to change, depending on the direction and extent of bias. Finally, the
effects of a bias will be augmented if the argument sample generated while
reconsidering is more exhaustive then the sample generated prior to initial
choice.
A number of situational factors guide (bias) the person in generating
arguments. For instance, knowing that a disagreement with the majority
will be made public may induce the individual to consider ideas that would
never have entered his mind if the disagreement had remained private. The
two sources of bias to be examined here involve the knowledge of other’s
preference for a particular alternative and the monopolization
of attention
by this alternative. Both influence the representativeness and exhaustiveness of an argument sample. This means that they determine the likelihood
of new ideas, and thus the likelihood of attitude change. We will illustrate
these processes in terms of some well known results from Asch (1951,1956)
having to do with the size of a majority and the nonunamity of a majority.
An important and still puzzling finding is that the tendency to agree with a
unaminous majority increased with the size of the majority but only up to
the size of three (Asch, 1956). Hence, a unaminous majority need not be
large to exert all of its potential influence. To begin, it is worth keeping in
mind that individual choices are made known in series. Thus, being next to
last, the person receives more information about the preferences of others
and has a greater amount of time to generate arguments as majority size
increases. The knowledge that another prefers, say, alternative X rather
than Y, focuses the person’s attention on the former; the larger the
majority, the more this alternative will monopolize attention and lead to the
generation of a disproportionately
large number of pro-X arguments.
Therefore, the tendency to agree with a majority should continue to
increase over a large range of sizes. Asch did not obtain such an effect and
persuasive-arguments
suggests why. At some point in the series the person
will have exhausted the pool of pro-X arguments. From that point on,
knowledge of another’s preference will have no impact; the person simply
will be unable to think of additional reasons for another choosing X.
Whether this point occurs early or late in the series obviously depends on
the number of pro-X arguments in the pool. In Asch’s study the alternative
preferred by the majority is unlikely to elicit many favorable arguments.
Hence, it is no surprise that under these conditions the tendency to
conform does not increase with the majority size beyond some relatively
small value.
The influence of a majority can be considerably diminished if it is less
than unanimous. The most common explanation of this effect assumes that
the dissenter “liberates” the person to behave as he would really like. The
implications of this analysis are rather different from those of persuasivearguments theory. For instance, the former suggests that once freed by the
ATTITUDE
POLARIZATION
329
example of dissent a person will act correctly, as he would in the absence of
an incorrect majority. From the point of view of persuasive argumentation,
however, the minority merely guides attention away from the majority
position, either in a correct or incorrect direction, depending on the
minority’s choice. Suppose the person is to select one of the three distinct
alternatives, X, Y, or Z; we know that X is correct (i.e., there exist many
highly persuasive arguments in its favor), Y is moderately incorrect (i.e.,
there exists a medium number of supporting arguments), and Z is extremely
incorrect (i.e., there exist few if any supporting arguments). Hence, when
individuals choose alone, there is an overwhelming preference for X. Now,
however, a majority selects Z while a minority prefers Y. If the presence of
a minority liberates the person to act as he normally would, then we should
observe not merely a weakening of conformity
(a decrease in the
preference for Z) but also a strengthening of correct behavior (an increase
in the preference for X). On the other hand, if a minority simply leads the
person to think more extensively about its choice, no change in preference
for the correct alternative would be observed. Instead, there should be a
decrease in conformity coupled with an increased preference for the
moderately incorrect alternative, Y. We know of only one relevant study:
Asch (1951) had a majority of seven choose the extremely incorrect line
while a minority of one chose the moderately incorrect line. His findings
were completely consistent with the persuasive-arguments
analysis. There
was a considerable shift in preference toward the moderately incorrect
alternative but no change in the preference for the correct alternative.
CONCLUSIONS
AND SUMMARY
According to Sanders and Baron “the question is not whether
comparison is sufficient to produce choice shifts under all conditions but
rather whether it is sufficient under the conditions in which choice shifts are
typically (our italics) obtained.” At least in respect to typical issues, which
involve rich ideation, past research indicates that social comparison is
neither a necessary nor a sufficient condition for polarization. At present, it
seems this phenomenon
can be adequately explained by persuasive
argumentation
alone. From a broader theoretical perspective, however,
the relationship between persuasive-arguments
and social comparison
might be conceptualized as one in which the former deals with the more
immediate
determinants of polarization
and the latter, with relatively
remote determinants. Finally, it was suggested that persuasive-arguments
could be extended beyond the typical polarization
situtation to those
involving ostensibly “undiscussable”
issues, where argumentation seems
virtually unimaginable
(e.g., arguments for the incorrect line in Asch).
Traditional
analyses of social influence say that under such conditions
changes in attitude must be due to normative processes. If this extension
BURNSTEIN
330
AND VINOKUR
has merit, it suggests rather that social influence in general is essentially
informational.
REFERENCES
Anderson, N. H., & Graesser, C. C. An information integration analysis of attitude change in
group discussion. Journal of Personality
and Social Psychology,
1976,34,210-222.
Asch, S. E. Effects of group pressure on the modification and distortion of judgments. In H.
Geutzkow (Ed.), Groups, leadership,
and men. Pittsburgh: Carnegie, 1951.
Asch, S. E. Studies of independence and conformities. Psychological
Monographs,
1956.70,
No. 9 (Whole No. 416).
Baron, R. S., & Roper, G. Reaffirmation of social comparison views of choice shifts:
Averaging and extremity effects in an autokinetic situation. Journal of Personality
and
Social Psychology,
1976, 33, 521-530.
Bateson, N. Familiarization, group discussion, and risk-taking. Journal of Experimental
Social Psychology,
1966, 2, 119- 129.
Bell, P. R., & Jamieson, B. D. Publicity of initial decisions and the risky shift phenomenon.
Journal
of Experimental
Social Psychology,
1970, 6, 329-345.
Blascovich, J., & Ginsburg, G. Blackjack and the risky shift. Sociometty,
1973, 36, 42-55.
Blascovich, J., Ginsburg, G., & Veach, T. A pluralistic explanation of choice shifts on the risk
dimenstion. Journal of Personality
and Social Psychology,
1975, 31, 422-429.
Brickman, P., Redfield, J., Harrison, A. A., & Crandall, R. Drive and predisposition as
factors in attitudinal effects of mere exposure. Journal
of Experimental
Social
Psychology,
1972, 8, 31-44.
Brown, R. Social Psychology.
New York: Free Press of Glencoe, 1965.
Bumstein, E., Miller, H., Vinokur, A., Katz, S., & Crowley, J. The risky shift is eminently
rational. Journal of Personality
and Social Psychology,
197 1, 20, 462-471.
Bumstein, E., & Vinokur, A. Testing two classes of theories about group-induced shifts in
individual choice. Journal of Experimental
Social Psychology,
1973, 9, 123- 137.
Bumstein, E., & Vinokur, A. What a person thinks upon learning he has chosen differently
from others: Nice evidence for the persuasive-arguments explanation of choice shifts.
Journal
of Experimental
Social Psychology,
1975, l&412-426.
Bumstein, E., Vinokur, A., & Trope, Y. Interpersonal comparison versus persuasive
argumentation: A more direct test of alternative explanations for group induced shifts
in individual choice. Journal of Experimental
Social Psychology,
1973, 9, 236-245.
Clark, R. D., & Willems, E. P. Where is the risky shift? Journal of Personality
and Social
Psychology,
1969, 13, 215-221.
Clark, R. D., Crockett, W. H., & Archer, R. L. Risk as value hypothesis: The relationship
between perception of self, others, and the risky shift. Journal of Personality
and
Social Psychology,
1971, 20, 425-429.
DesJarlais, D. Shifts to risk and reward. Psychological
Reports,
1970, 26, 119- 122.
Ferguson, D. A., & Vidmar, N. Effects of group discussion on estimates of culturally
appropriate risk levels. Journal of Personality
and Social Psychology,
1971, 20,
436-445.
Festinger, L. A theory of social comparison processes. Human Relations,
1954,7,117- 140.
Flanders, J. P., & Thistlethwaite, D. L. Effects of familiarization and group discussion upon
risk-taking. Journal of Personality
and Social Psychology,
1967, 5, 91-98.
Gush, J. E. Attitude formation and mere exposure: A non-artifactual explanation of empirical
findings. Journal of Personality
and Social Psychology,
1976. 33,281-290.
Jellison, M., & Riskind, J. A social comparison of abilities interpretation of risk taking
behavior. Journal of Personality
and Social Psychology,
1970, 15, 375-390.
Kahan, J. P. A subjective probability interpretation of the risky shift. Journal ofPersonality
and Social Psychology,
1975. 31, 977-982.
ATTITUDE
POLARIZATION
331
Kaplan, M. F. Judgment by juries. In M. F. Kaplan & S. Schwartz (Eds.), Human judgment
and decision-making
in applied settings.
New York: Academic Press, 1977.
Lamm, H., Trommsdorff, G., & Rost-Schaude, E. Self-image, perception of peers’ risks
acceptance and risky shift. European Journal of Social Psychology,
1972,2,255-272.
Levinger, G., & Schneider, D. J. A test of the “Risk is a Value” hypothesis.
Journal
of Personality
and Social Psychology,
11, 165- 169.
Myers, D. G., & Lamm, H. The group polarization phenomenon. Psychological
Bulletin,
1976, 83, 602-627.
Psychological
Bulletin,
1970, 73,
Natsoulas, T. Concerning introspective “knowledge.”
89-111.
Perlman, D., & Oskamp, S. The effects of picture content and exposure frequency on
evaluations of Negroes and whites. Journal of Experimental
Social Psycology
,197l. 7,
503-514.
Pruitt, D. G. Choice shifts in group discussion: An introductory review. Journal
of
Personality
and Social Psychology,
1971, 20, 339-360.
Pylyshyn, Z. W. What the mind’s eye tells the mind’s brain: A critique of mental imagery.
Psychological
Bulletin,
1973, 80, l-24.
Sadler, O., & Tesser, A. Some effects of salience and time upon interpersonal hostility and
attraction during social isolation. Sociometry
, 1973, 36, 99- 112.
Sanders, G., & Baron, R. S. Is social comparison irrelevant for producing choice shifts?
Journal
of Experimental
Social Psychology,
1977, 13,303-314.
St. Jean, R. Reformation of the value hypothesis in group risk taking. Proceeding
of the 78th
Annual
Convention
of the American
Psychological
Association,
1970, 5, 339-340.
Stokes, J. P. Effects of familiarization and knowledge of others’ odd choices on shifts to risk
and caution. Journal of Personality
and Social Psychology,
1971, 20, 407-412.
Teger, A. I., & Pruitt, D. G. Components of group risk taking. Journal ofExperimental
Social
Psychology,
1967, 3, 189-205.
Teger, A. I., Pruitt, D. G., St. Jean, R. & Haaiand, G. A. A re-examination of the
familiarization hypothesis in group risk taking. Journal
of Experimental
Social
Psychology,
1970, 6, 346-350.
Tesser, A. Attitude polarization as a function of thought and reality constraints. Journal of
Research
in Personality,
1976, 10, 183- 194.
Tesser, A., & Conlee, M. C. Some effects of time and thought on attitude polarization.
Journal
of Personality
and Social Psychology,
1975, 31, 262-270.
Tesser, A., & Cowan, C. L. Some effects of thought and number of cognitions on attitude
change. Social Behavior and Personality,
in press.
Tesser, A., & Leone, C. Cognitive schema and thought as determinants of attitude change.
Journal
of Experimental
Social Psychology,
in press.
Tuddenham, R. D., & McBride, P. D. The yielding experiment from the subject’s point of
view. Journal of Personality,
1959, 27, 259-271.
Vinokur, A. Review and theoretical analysis of the effects of group processes upon individual
and group decisions involving risk. Psychological
Bulletin,
1971a, 76, 231-250.
Vinokur, A. Cognitive and affective processes influencing risk-taking in groups: An expected
utility approach. Journal of Personlaity
and Social Psychology,
1971b, 20,472-486.
Vinokur, A., & Bumstein, E. The effects of partially shared persuasive agruments on group
induced shifts: A group problem solving approach. Journal of Personality
and Social
Psychology,
1974, 29, 305-315.
Vinokur, A., & Burnstein, E. Novel argumentation and attitude change: The case of
polarization following group discussion. European
Journal
of Social Psychology,
in
press.
Vinokur, A., Trope, Y., & Bumstein, E. A decision-making analysis of persuasive
argumentation and the choice-shift effect. Journal OfExperimental
Social Psychology,
1975, 11, 127-148.
BURNSTEIN
332
AND VINOKUR
Wallach, M. A., & Kogan, N. The roles of information, discussion, and consensus in group
risk taking. Journal
of Experimental
Social Psychology,
1965, 1, I-19.
Zajonc, R. B. The attitudinal effects of mere exposure. Journal of Personality
and Social
Psychology,
1968, 9, l-27.
REFERENCE
1. Kaplan,
M. F. Group
discussion
effects
NOTES
in
a modified
jury
decision
paradigm:
Paper presented at the meeting of the Psychonomic Society,
Denver, November 1976.
2. Kaplan, M. F., & Miller, C. E. Juror judgments
and discussion:
Effects of variety
of
shared information
on amount of polarization.
Paper presented at the meeting of the
Psychonomic Society, St. Louis, November 1976.
3. St. Jean, R., & Percival, E. The effects of information
control on shifts in risk taking.
Paper presented at the Eastern Psychological Association, May 1973.
Information
influences.